Method and apparatus for adaptive frame prediction
Abstract
An adaptive frame prediction method includes: using at least one block partition method to partition an MB that needs to perform intra-frame prediction; using at least one prediction mode to predict each sub-block in each block partition result and obtaining a sub-block residual of a current sub-block; selecting multiple transformation methods according to the sub-block type to transform each sub-block residual and obtain the code-rate distortion ratio of the current sub-block; performing frame prediction for the MB according to the block partition method, the prediction mode, and the transformation method corresponding to the code-rate distortion ratio of a minimum value among all code-rate distortion ratios. An adaptive prediction apparatus disclosed herein includes: a block partition module, a predicting module, a transforming module, a selecting module, and a frame predicting module. The method and the apparatus herein perform the intra-frame or inter-frame coding flexibly, get more adaptive to the local texture statistic features of the MBs, and improve the coding efficiency.
Claims
exact text as granted — not AI-modified1 . An adaptive frame prediction method, comprising:
using at least one block partition method to partition a Macro Block (MB) that needs to perform intra-frame prediction; using at least one prediction mode to predict each sub-block in each block partition result and obtaining a sub-block residual of a current sub-block; using at least one transformation method to transform each sub-block residual; obtaining a code-rate distortion ratio of the current sub-block according to a type of the sub-block, a current prediction mode, and the transformation method; and performing frame prediction for the MB according to a block partition method, a prediction mode, and a transformation method corresponding to a code-rate distortion ratio of a minimum value among all code-rate distortion ratios.
2 . The adaptive frame prediction method of claim 1 , wherein:
if the frame prediction is intra-frame prediction, the block partition method comprises at least one irregular block partition method.
3 . The adaptive frame prediction method of claim 2 , wherein the irregular block partition methods comprise:
a block partition method which partitions each rectangular sub-block into two non-rectangular sub-blocks along a straight line of non-vertical and non-horizontal.
4 . The adaptive frame prediction method of claim 3 , wherein the irregular block partition methods comprise:
a block partition method which partitions each rectangular sub-block into two non-rectangular sub-blocks along a diagonal line of a rectangle.
5 . The adaptive frame prediction method of claim 2 , wherein the using of the at least one transformation method to transform each sub-block residual comprises:
if a type of the current sub-block is a sub-block generated through the irregular block partition method, using the at least one transformation method except an Integer Discrete Cosine Transformation (IDCT) method to transform the sub-block residual of the current sub-block; and if the type of the current sub-block is a sub-block generated through a regular block partition method, using at least two transformation methods to transform the sub-block residual of the current sub-block.
6 . The adaptive frame prediction method of claim 1 , wherein the transformation methods comprise: Integer Discrete Cosine Transformation (IDCT), Karhunen-Loeve (K-L) transformation, and wavelet transformation.
7 . The adaptive frame prediction method of claim 3 , wherein when using Karhunen-Loeve (K-L) transformation to transform the sub-block residual:
searching a preset unified K-L transformation code table for a best K-L transformation coefficient according to different block partition methods and different prediction modes, and performing the K-L transformation for the sub-block residual according to the best K-L transformation coefficient; or obtaining corresponding K-L transformation coefficient code tables respectively according to different block partition methods and different prediction modes, and performing the K-L transformation for the sub-block residual according to corresponding K-L transformation coefficients.
8 . The adaptive frame prediction method of claim 1 , further comprising:
setting a block partition flag bit in MB header information, and determining the block partition method for partitioning the MB according to a type of the MB and a value of the block partition flag bit.
9 . The adaptive frame prediction method of claim 1 , further comprising:
setting a transformation flag bit after MB prediction information, and determining a transformation method for transforming the sub-block residual of the current sub-block according to a value of the transformation flag bit.
10 . The adaptive frame prediction method of claim 1 , wherein the obtaining of the code-rate distortion ratio of the current sub-block according to the type of the sub-block, the current prediction mode, and the transformation method comprises:
selecting at least one transformation method according to a preset type of the sub-block and the current prediction mode, and obtaining the code-rate distortion ratio of the current sub-block; or selecting any one of types of the sub-blocks as a combined type; selecting any one of current prediction modes as a combined mode; selecting any one of the transformation methods as a combined method; and obtaining the code-rate distortion ratio of the current sub-block according to the combined type, the combined mode, and the combined method.
11 . An adaptive frame prediction apparatus, comprising:
a block partition module, adapted to use at least one block partition method to partition a Macro Block (MB) that needs to perform intra-frame prediction, and send a block partition result to a predicting module; the predicting module, adapted to use at least one prediction mode to predict each sub-block in each received block partition result, obtain a sub-block residual of a current sub-block, and send the sub-block residual to a transforming module; the transforming module, adapted to: use at least one transformation method to transform each sub-block residual; obtain a code-rate distortion ratio of the current sub-block according to a type of the sub-block, a current prediction mode, and the transformation method; and send the code-rate distortion ratio to a selecting module; the selecting module, adapted to select a code-rate distortion ratio of a minimum value among all obtained code-rate distortion ratios, and send the selected code-rate distortion ratio to a frame predicting module; and the frame predicting module, adapted to perform frame prediction for the MB according to a block partition method, a prediction mode, and a transformation method corresponding to the received code-rate distortion ratio.Join the waitlist — get patent alerts
Track US2010195715A1 — get alerts on status changes and closely related new filings.
We store only your email — no account needed. See our privacy policy.